From 996a4e2f425e67930f612c038e57db875329eb25 Mon Sep 17 00:00:00 2001
From: David Ndungu <ndungu.sink@gmail.com>
Date: Wed, 15 Apr 2026 19:34:36 -0700
Subject: [PATCH] feat(graph): T99.1.2 mark Gemma4PLECombinedProducer
 non-capturable

Gemma4PLECombinedProducer performs a CPU-side gather over the shared
PLE embedding table and then calls MulScalar on the freshly-allocated
CPUStorage tensor. Inside a CUDA graph capture stream this triggers a
synchronous H2D cudaMemcpy that CUDA rejects with "operation would make
the legacy stream depend on a capturing blocking stream".

Add the op to nonCapturableOps so the producer runs in pre-capture on
every forward, outside the capturing stream. The producer runs once
per forward pass before the transformer loop, so this placement keeps
the layer-body capture region intact.

Companion change in zerfoo/inference/gemma4_edge_ple_nodes.go
(E99 T99.1.2) pre-slices the producer's outputs into stable GPU
buffers so pleSliceNode stays fully capturable.

Decision recorded in zerfoo/docs/adr/088-gemma4-ple-cuda-graph-capture.md.
---
 graph/cuda_graph.go | 22 +++++++++++++++-------
 1 file changed, 15 insertions(+), 7 deletions(-)

diff --git a/graph/cuda_graph.go b/graph/cuda_graph.go
index 009aedf..f813dbb 100644
--- a/graph/cuda_graph.go
+++ b/graph/cuda_graph.go
@@ -44,14 +44,22 @@ var debugGraphCapture = os.Getenv("ZERFOO_DEBUG_GPU") == "1"
 // (offset_memcpy kernel) and GQA uses GPU RoPE selection (rope_select kernel),
 // all position-dependent state is read from GPU memory at replay time, making
 // GQA fully capturable.
+//
+// Gemma4PLECombinedProducer: performs a CPU-side gather over the shared PLE
+// embedding table (token ids -> per-layer rows), then calls MulScalar on the
+// freshly-allocated CPUStorage tensor. Running this inside a capture stream
+// triggers a synchronous H2D cudaMemcpy that CUDA rejects. The producer runs
+// once per forward pass before the transformer loop, so placing it in
+// pre-capture keeps the layer-body capture region intact. See ADR-088.
 var nonCapturableOps = map[string]bool{
-	"EmbeddingLookup":   true,
-	"Gather":            true,
-	"AutoAttentionMask": true,
-	"AutoPositionIds":   true,
-	"Slice":             true,
-	"ConstantOfShape":   true,
-	"Shape":             true,
+	"EmbeddingLookup":            true,
+	"Gather":                     true,
+	"AutoAttentionMask":          true,
+	"AutoPositionIds":            true,
+	"Slice":                      true,
+	"ConstantOfShape":            true,
+	"Shape":                      true,
+	"Gemma4PLECombinedProducer":  true,
 }
 
 // isNonCapturable returns true if the instruction at index i in the plan